Linear Process Bootstrap Unit Root Test

TL;DR

This paper introduces a bootstrap-based unit root test designed to address size distortions caused by moving average noise, improving upon the Phillips-Perron test while maintaining high power.

Contribution

It proposes a nonparametric bootstrap method for unit root testing that is consistent for a wide range of non-linear time series with moving average noise.

Findings

Bootstrap test reduces size distortions compared to Phillips-Perron.

Maintains high power in simulations.

Proven consistency via bootstrap functional central limit theorem.

Abstract

One of the most widely applied unit root test, Phillips-Perron test, enjoys in general highpowers, but suffers from size distortions when moving average noise exists. As a remedy, thispaper proposes a nonparametric bootstrap unit root test that specifically targets moving aver-age noise. Via a bootstrap functional central limit theorem, the consistency of this bootstrapapproach is established under general assumptions which allows a large family of non-linear timeseries. In simulation, this bootstrap test alleviates the size distortions of the Phillips-Perrontest while preserving its high powers.